Classifying means of transportation using mobile sensor data

Mobile phone sensors have the ability to provide significant information about environmental conditions as well as about different activities of persons. This paper deals with the acceleration sensor of a commercial mobile phone that makes it possible to identify and classify the user's means of transportation. The needed pre-processing of the sensor data is described as well as the used classification algorithms, i.e. Naive Bayes classifier and Support Vector Machine (SVM). Their performance for solving the classification problem in combination with the different pre-processings is analyzed. It is shown that both classifiers are able to solve the classification task with high accuracy given a proper preprocessing. Support Vector Machines outperform Naive Bayes classifiers and achieve a classification accuracy of over 97% on an unknown test data set.